Otoscopic diagnosis using computer vision: An automated machine learning approach
Bibliographic record
Abstract
OBJECTIVE: Access to otolaryngology is limited by lengthy wait lists and lack of specialists, especially in rural and remote areas. The objective of this study was to use an automated machine learning approach to build a computer vision algorithm for otoscopic diagnosis capable of greater accuracy than trained physicians. This algorithm could be used by primary care providers to facilitate timely referral, triage, and effective treatment. METHODS: Otoscopic images were obtained from Google Images (Google Inc., Mountain View, CA), from open access repositories, and within otolaryngology clinics associated with our institution. After preprocessing, 1,366 unique images were uploaded to the Google Cloud Vision AutoML platform (Google Inc.) and annotated with one or more of 14 otologic diagnoses. A consensus set of labels for each otoscopic image was attained, and a multilabel classifier architecture algorithm was trained. The performance of the algorithm on an 89-image test set was compared to the performance of physicians from pediatrics, emergency medicine, otolaryngology, and family medicine. RESULTS: For all diagnoses combined, the average precision (positive predictive value) of the algorithm was 90.9%, and the average recall (sensitivity) was 86.1%. The algorithm made 79 correct diagnoses with an accuracy of 88.7%. The average physician accuracy was 58.9%. CONCLUSION: We have created a computer vision algorithm using automated machine learning that on average rivals the accuracy of the physicians we tested. Fourteen different otologic diagnoses were analyzed. The field of medicine will be changed dramatically by artificial intelligence within the next few decades, and physicians of all specialties must be prepared to guide that process. LEVEL OF EVIDENCE: NA Laryngoscope, 130:1408-1413, 2020.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.005 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.003 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".